| Autor: |
Yu, Lian-Hui, Li, Xiao-Yu, Chen, Geng, Zhu, Qin-Sheng, Li, Hui, Yang, Guo-Wu |
| Rok vydání: |
2024 |
| Předmět: |
|
| Druh dokumentu: |
Working Paper |
| Popis: |
Quantum architecture search (QAS) represents a fundamental challenge in quantum machine learning, with SOTA methods primarily categorized into training-free and gradient-guided approaches. However, treating QAS solely as either a discrete pruning process or a continuous optimization problem fails to balance accuracy and efficiency. This work decomposes QAS into two alternately solved subproblems: optimal circuit structure retrieval and parameter optimization. Based on this insight, we propose Quantum Untrained-Explored Synergistic Trained Architecture (QUEST-A), which implements rapid architecture pruning through inherent circuit properties and develops focused optimization with parameter reuse strategies. QUEST-A unifies discrete structure search and continuous parameter optimization within an evolutionary framework that integrates rapid pruning and fine-grained optimization. Experiments demonstrate QUEST-A's superiority over existing methods: enhancing model expressivity in signal representation, maintaining high performance across varying complexities in image classification, and achieving order-of-magnitude precision improvements in variational quantum eigensolver tasks. These results validate QUEST-A's effectiveness and provide transferable methodologies for QAS. |
| Databáze: |
arXiv |
| Externí odkaz: |
|
